Integrative analysis and external validation identify a novel seven-gene signature for preeclampsia risk stratification

Placental transcriptomic alterations play a critical role in the pathophysiology of preeclampsia (PE), yet robust gene signatures for risk stratification remain underexplored. We first integrated four GEO datasets (GSE75010, GSE25906, GSE24129, and GSE10588). Using the GSE75010 dataset as the training cohort, we applied LASSO regression and SVM algorithms to screen and identify seven key genes. A risk stratification model for PE was then constructed based on these seven genes. The model’s performance was validated in the other three independent GEO datasets. Finally, transcriptomic sequencing was conducted on placental tissues from 10 PE patients and 10 controls, and qPCR and immunohistochemistry (IHC) were performed on an independent set of 5 PE patients and 5 controls, to experimentally validate the expression of the key genes. We identified seven key genes (ACOXL, FADS2, HPGD, CPOX, LDHA, BMPR1B, GAPDHS) using intersection of multiple machine learning algorithms. The risk model achieved an AUC of 0.893 for discriminating PE from normal controls. ACOXL and FADS2 were highly expressed in the low-risk group, while the remaining five genes were elevated in the high-risk group, with consistent expression patterns across three external datasets. WGCNA revealed that the high-risk group was primarily enriched in hypoxia and extracellular matrix adhesion pathways. The nomogram showed good performance with AUCs of 0.945, 0.882, 0.847, and 0.836 in the training and three validation sets, respectively. Experimental validation in our placental cohort confirmed the consistency of these gene expressions.